AI developer xAI leveraging Anthropic’s Claude AI model through user accounts after restricted access, illustrating tech indu

Editorial illustration for xAI used Anthropic’s Claude via personal accounts after access revoked for months

xAI used Anthropic’s Claude via personal accounts after...

xAI used Anthropic’s Claude via personal accounts after access revoked for months

By AI Daily Post Edited by Brian Petersen, Editor-in-Chief

June 6, 2026 • 2 min read

Elon Musk’s xAI has been quietly leaning on Anthropic’s Claude to shape its own coding models. According to a report from The Information, the startup fed Claude’s output directly into its pre‑training pipeline for months. When Anthropic pulled official access in January, xAI didn’t stop; engineers rerouted the data through personal accounts and an intermediary service called Blackbox AI.

The move mirrors Musk’s earlier courtroom admission that Grok was “partially” trained on OpenAI models, a practice he called industry standard. Inside xAI, the picture looks strained: the pre‑training team has shrunk to fewer than five people, four code leads have departed within months, and a single employee’s accidental deletion of training data erased two to three weeks of work. Meanwhile, the compute hardware Musk bought is now being leased to Anthropic via SpaceX and to Google, rather than powering in‑house model development.

The arrangement appears to be a stopgap, but it raises questions about how reliant xAI is on external AI outputs to keep its own products moving forward.

After Anthropic revoked official access in January, xAI engineers kept going through personal accounts and the intermediary service Blackbox AI. Musk previously admitted in court that xAI "partially" used OpenAI models to train Grok, calling it industry standard. The pretraining team shrank to under five people.

Four Grok code leads left within months, along with many co-founders. One employee accidentally deleted critical training data, costing two to three weeks of work, according to The Information. He's now renting it to Anthropic via SpaceX and to Google instead of training his own models.

Elon Musk's xAI reportedly trained its coding models on Claude outputs for months before getting cut off - THE DECODER

Why this matters

Developers should note that xAI’s coding models were built on Claude’s outputs for months, even after Anthropic cut off official access in January. How sustainable that approach is remains unclear; the team resorted to personal accounts and the Blackbox AI intermediary to keep the pipeline alive. Musk’s court admission that Grok was “partially” trained on OpenAI models adds another layer of complexity, suggesting a blend of external data sources that he describes as “industry standard.” Yet the shrinking of the pre‑training team to under‑ten engineers raises questions about the depth of expertise overseeing this integration.

For founders, the episode underscores the risk of relying on third‑party models without solid licensing agreements. Researchers may wonder whether the distilled knowledge from Claude retains fidelity or introduces hidden biases. We must watch how xAI addresses these uncertainties, especially if similar practices become common across the sector.

The broader implication: transparency and clear provenance of training data are becoming essential checkpoints for any AI venture.

xAI used Anthropic’s Claude via personal accounts after...

Further Reading

Latest News

Study Finds No One-Size-Fits-All Strategy for Multi-Agent Communication

Reddit releases AI comment archive to study LLM persuasion tactics